Optimize Ernie 4.5 VL timestamp rendering with cached overlays

[raimbekovm]

What does this PR do?

Optimizes timestamp rendering in Ernie 4.5 VL video processing by caching text overlays and using torch alpha blending instead of slow torch->PIL->torch conversion for each frame.

Performance improvement:

Mode	Time/frame	Speedup
Original	~2.2 ms	1x
Cold cache	1.5 ms	1.5x
Warm cache	0.17 ms	~13x

For 180 frames (default max): ~400ms → ~30ms

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline, Pull Request section?
• Was this discussed/approved via a Github issue or the forum?
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

Who can review?

@yonigozlan @molbap

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: ernie4_5_vl_moe